summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorSubv <subv2112@gmail.com>2018-04-25 05:00:40 +0200
committerSubv <subv2112@gmail.com>2018-04-25 19:54:26 +0200
commit20d86d8a36dca4bf1b465193745a365c3ab9abcd (patch)
tree632e5bf883d27a84d4891bad54ac1426023097ed
parentMemory: Added a missing shortcut for Memory::CopyBlock for the current process. (diff)
downloadyuzu-20d86d8a36dca4bf1b465193745a365c3ab9abcd.tar
yuzu-20d86d8a36dca4bf1b465193745a365c3ab9abcd.tar.gz
yuzu-20d86d8a36dca4bf1b465193745a365c3ab9abcd.tar.bz2
yuzu-20d86d8a36dca4bf1b465193745a365c3ab9abcd.tar.lz
yuzu-20d86d8a36dca4bf1b465193745a365c3ab9abcd.tar.xz
yuzu-20d86d8a36dca4bf1b465193745a365c3ab9abcd.tar.zst
yuzu-20d86d8a36dca4bf1b465193745a365c3ab9abcd.zip
-rw-r--r--src/video_core/engines/fermi_2d.cpp54
-rw-r--r--src/video_core/engines/fermi_2d.h5
2 files changed, 59 insertions, 0 deletions
diff --git a/src/video_core/engines/fermi_2d.cpp b/src/video_core/engines/fermi_2d.cpp
index 87634da21..9019f2504 100644
--- a/src/video_core/engines/fermi_2d.cpp
+++ b/src/video_core/engines/fermi_2d.cpp
@@ -2,7 +2,9 @@
// Licensed under GPLv2 or any later version
// Refer to the license.txt file included.
+#include "core/memory.h"
#include "video_core/engines/fermi_2d.h"
+#include "video_core/textures/decoders.h"
namespace Tegra {
namespace Engines {
@@ -12,6 +14,58 @@ Fermi2D::Fermi2D(MemoryManager& memory_manager) : memory_manager(memory_manager)
void Fermi2D::WriteReg(u32 method, u32 value) {
ASSERT_MSG(method < Regs::NUM_REGS,
"Invalid Fermi2D register, increase the size of the Regs structure");
+
+ regs.reg_array[method] = value;
+
+ switch (method) {
+ case FERMI2D_REG_INDEX(trigger): {
+ HandleSurfaceCopy();
+ break;
+ }
+ }
+}
+
+void Fermi2D::HandleSurfaceCopy() {
+ NGLOG_WARNING(HW_GPU, "Requested a surface copy with operation {}",
+ static_cast<u32>(regs.operation));
+
+ const GPUVAddr source = regs.src.Address();
+ const GPUVAddr dest = regs.dst.Address();
+
+ // TODO(Subv): Only same-format and same-size copies are allowed for now.
+ ASSERT(regs.src.format == regs.dst.format);
+ ASSERT(regs.src.width * regs.src.height == regs.dst.width * regs.dst.height);
+
+ // TODO(Subv): Only raw copies are implemented.
+ ASSERT(regs.operation == Regs::Operation::SrcCopy);
+
+ const VAddr source_cpu = *memory_manager.GpuToCpuAddress(source);
+ const VAddr dest_cpu = *memory_manager.GpuToCpuAddress(dest);
+
+ u32 src_bytes_per_pixel = RenderTargetBytesPerPixel(regs.src.format);
+ u32 dst_bytes_per_pixel = RenderTargetBytesPerPixel(regs.dst.format);
+
+ if (regs.src.linear == regs.dst.linear) {
+ // If the input layout and the output layout are the same, just perform a raw copy.
+ Memory::CopyBlock(dest_cpu, source_cpu,
+ src_bytes_per_pixel * regs.dst.width * regs.dst.height);
+ return;
+ }
+
+ u8* src_buffer = Memory::GetPointer(source_cpu);
+ u8* dst_buffer = Memory::GetPointer(dest_cpu);
+
+ if (!regs.src.linear && regs.dst.linear) {
+ // If the input is tiled and the output is linear, deswizzle the input and copy it over.
+ Texture::CopySwizzledData(regs.src.width, regs.src.height, src_bytes_per_pixel,
+ dst_bytes_per_pixel, src_buffer, dst_buffer, true,
+ regs.src.block_height);
+ } else {
+ // If the input is linear and the output is tiled, swizzle the input and copy it over.
+ Texture::CopySwizzledData(regs.src.width, regs.src.height, src_bytes_per_pixel,
+ dst_bytes_per_pixel, dst_buffer, src_buffer, false,
+ regs.dst.block_height);
+ }
}
} // namespace Engines
diff --git a/src/video_core/engines/fermi_2d.h b/src/video_core/engines/fermi_2d.h
index 78d06218f..0c5b413cc 100644
--- a/src/video_core/engines/fermi_2d.h
+++ b/src/video_core/engines/fermi_2d.h
@@ -88,6 +88,11 @@ public:
} regs{};
MemoryManager& memory_manager;
+
+private:
+ /// Performs the copy from the source surface to the destination surface as configured in the
+ /// registers.
+ void HandleSurfaceCopy();
};
#define ASSERT_REG_POSITION(field_name, position) \